Natural Hazards Research Summit 2024: Unstructured to Actionable: Extracting Wind Event Impact Data for Enhanced Infrastructure Resilience

This project employs zero-shot text classification with pre-trained BART-large models to efficiently extract and analyze critical data on infrastructure and community impacts from wind disaster reconnaissance reports. While primarily focused on wind, the methodology can be applied to various hazards, including earthquakes. It utilizes advanced NLP models including BART-large MNLI and CNN, which eliminates the need for extensive labeled datasets. This approach enables the rapid synthesis of impact information from historical damage reports, essential for informed decision-making and resilience planning. The primary audience includes researchers, disaster managers, and natural hazard engineers, focusing on disaster resilience and response.
For more details and access to the project resources, visit the GitHub repository: Impact-Data-Mining.